It’s time for an entry in my weekly coding journal.
-= Spring Boot =-
I watched some videos, read the tutorials, and started building a Spring Boot application. I planned to create an app that tracks page visits for my personal website. I intend to store the visit count for each URL in an SQLite database. For database access, I’m using Java Persistence API from the spring-boot-starter-data-jpa library.
I’ve already configured Docker and Kubernetes for the new application in my cluster. The Spring Boot app uses about 100…200 megabytes of RAM, so I allocated a 300 MB limit to it.
Last week, I spent 9 hours studying Spring Boot. I started reluctantly, but gradually managed to get into the groove.
-= SmartProgress =-
In one of my previous posts, I wrote that SmartProgress enabled Cloudflare bot protection, which blocked my blog post automation. Well, that protection has been gone for a couple of weeks now 🥴 My automation is back in action. I still plan to make my personal website the primary source for my posts, but I’ll do that sometime later if the right moment presents itself.
-= AI Video Generator =-
I set up an AI video generator on my home computer. Previously, I had a Stable Diffusion static image generator and an LLM for text generation. Now, I’ve set up the Wan 2.2 14B model. It took some fiddling… There is good news and bad news.
- Good News 🚀 The generator works, and it’s one of the most powerful video generators available today… at least among those accessible to mere mortals on personal PCs. Of course, there are even more advanced models, but they require server-level hardware.
- Bad News 🐌 The generator is very slow. On my PC, creating a 6-second video takes about 6 hours 🥴 It’s impossible to wait in real-time, so I do this: I start the generator and go to sleep; by morning, the video is ready.
The model requires about 21 GB of VRAM, and I only have 20 GB. That “last gigabyte” kills all the speed because it doesn’t fit into the memory, causing an endless 🔄 cycle of data transfer between DDR4 and VRAM over the PCI bus.
Right now, I’m gradually fine-tuning the cfg, shift, and steps parameters. I think I’ll start posting stupid AI memes on some social network to see how it goes… Once I find all the right parameters, things will move quickly: I could churn out one micro-video a day, but I haven’t reached that point yet; I’m still gradually setting up the process ⚙️
